Odin dev-2026-11 Adds Android Support and Overhauls Compression Libraries
The latest Odin release adds Android compile targets, accelerates core compression libraries, and improves compiler memory efficiency and analysis.
Odin dev-2026-11 adds Android as a supported compile target via the new -subtarget:android option covering arm64, arm32, amd64, and i386 at API level 29 or higher. You can use odin bundle android to produce a signed APK or AAB, running aapt2, adding a prebuilt classes.dex, aligning the package, and signing it. The release includes Android bindings in core:sys/android and vendor:android, an example app in examples/android, and a CI job for all four ABIs. Android target triples now require a vendor name to prevent LLVM ABI errors, and links use --no-undefined.
The core:compress library was overhauled with Deflate support and a rewritten Inflate. Tested on a 5950X and an M4 Pro using the Silesia corpus, Inflate runs 2.1–2.2x faster than the old core version and beats vendor:zlib 1.3.2. Deflate levels 1–9 run faster than zlib 1.3.2, and level 10 produces output 0.64% smaller than zlib level 9 at about 1.4x the time. The update also adds core:compress/lz4, core:image/bng, PNG saving, and a rewritten JPEG decoder with progressive JPEG support.
Breaking changes include core:fmt padding widths with spaces unless the 0 flag is given and width specifiers counting runes instead of bytes. In core:compress, Image.pixels and Context.output are now [dynamic]byte, and image savers take ^[dynamic]byte. Packages core:compress/zlib and gzip are now deprecated wrappers, and several items like old LSB bit-reader helpers were removed.
Escape analysis for stack memory is now flow-based and interprocedural, while atomic ordering analysis is on by default and checks memory orderings across the whole program. Debug info is faster to generate using DWARF 5 on Darwin and Linux, automatically loading gdb and lldb pretty printers. Checker memory use dropped, with Scope shrinking from 568 to 104 bytes and odin check using about 16.5% less memory on large projects.